Research on the issue diversity and specialisation of cities over the last decade has shed new light on the "empirical regularity" of a relationship between city size and industrial diversity, identified in the early 1970s.Nevertheless, the results obtained from the theoretical models are not yet conclusive and indeed depend on the assumptions employed.Moreover, empirical research on the topic is relatively undeveloped and the stylised facts are not yet fully understood.In this paper we employ a panel data analysis for a big number (333) of US cities over a period of twenty-nine years , revealing two interesting results.First, time and city effects are highly significant.Second, while a positive relationship between size and diversity is apparent in the cross-sectional analysis, the time series and pooled investigations reveal a negative relationship.This, we interpret as evidence of the existence of a system of cities.In support to that, overall, our results suggest the existence of a complexity of mechanisms, which cannot be attributed simply to an exogenous city size effect.Under this perspective, the current theoretical advances in new economic geography seem to be the most useful tools in understanding the patterns and relationships identified in this paper.